Investigate alternative improvements in pgtextindex ranking

[Paul Everitt]

We have discussed ways to dramatically alter the relevance based on information outside the title/text. Examples:

- If someone tags a resource with "password" and a search is done with a searchterm that stems/synonyms to "password", give it a higher relevance than a word matching in the title

- A profile for someone who is a staff member should show up before a profile for an affiliate

- A resource authored by a staff member should show up before a resource authored by an affiliate

- Some tags (e.g. "london") could be promoted to "managed tags" aka topics which give them an even higher score

All of these are hypothetical. We will likely take whatever is reasonable with the machinery. Some machinery which could be investigated:

- Reflecting tag information into a column for "tags" similar to "title" and "text", and just giving that input source a higher weight during indexing (just like title does currently)

- Using more than one index and finding a way to normalize ranking across results, perhaps using a server-side function

- Changing the ranking functions as hinted at in "Different applications might require additional information for ranking, e.g., document modification time. The built-in ranking functions are only examples. You can write your own ranking functions and/or combine their results with additional factors to fit your specific needs."

[Paul Everitt (paul-agendaless)]

See the various chapter sections (prev/next) at http://durak.org/sean/pubs/software/postgresql/textsearch-controls.html